We both tried GrokBot. We disagree.

Over the past week or so, GrokBot has been all over X, in much the same way OpenClaw was earlier in the year. Some people are claiming they've cancelled their Claude subscriptions for it. We try to be anti-hype at Create With, so we've both been putting it to work. Kieran has been using it daily for about six days, James has been testing it too, and we came away with quite different experiences. This week you get both takes.

GrokBot is a web and mobile app that lets you build a team of AI bots that can actually do things for you. It sits in the same family as OpenClaw and Hermes, with a couple of differences that seem to have made it spread so fast.

Every bot you create gets its own virtual computer in the cloud. Think of each bot as a remote worker who turns up with their own laptop, and you can watch their screen at any time or grab the mouse and take over.

The bots can also message each other without you in the loop, so a group of them starts to feel like a small office rather than a row of separate chat windows.

Pricing has been chaotic. The entry price went from $200 a month to $60 to $30 in the space of about three days, so treat any number you see as a snapshot.

After six days of daily use, Kieran is impressed. The user experience is noticeably better than OpenClaw and better than Hermes, which was already easy to get started with.

His team so far:

• Frank, a chief of staff bot, manages all the other bots. Having one manager bot seems to be the accepted best practice, and it works well.
• Dee, a marketing bot, is connected to Google Search Console and runs regular SEO audits of his sites.
• A dev team picks up Dee's recommendations and implements them. They're connected to his GitHub and raise pull requests (packaged-up code changes) for him to approve.
• An inbox triage bot watches all his email accounts and briefs him on anything important each morning.
• A startup ideas team monitors trends using the X plugin, bounces ideas around between themselves, and only sends him the good ones.

The bots genuinely coordinate between themselves. Below, Charlie from the dev team and Frank settle a rollback question entirely on their own.

The feature Kieran would single out is multi-account support. He has several Google accounts, and GrokBot connected all of his Gmail and calendar accounts painlessly.

Claude has never handled this well. Connecting one Google account is easy, but juggling several has always been a pain, so if you run your life across multiple inboxes this matters more than it sounds.

The mobile app is decent too. It stays in sync with the desktop app, which is a nicer experience than using the Claude mobile app with Claude Code.

Some people have taken this much further. They've spun up huge numbers of bots, organised into teams, and effectively created a company structure that runs their personal life and multiple businesses at once.

Some of these setups include daily stand-up meetings, where the team manager bots discuss between themselves what their teams have been doing and how they could do better in future.

Whether this is genuinely an effective way to run GrokBot remains to be seen. But it is fascinating, and undeniably fun, to be able to spin up all these tiny AI employees so effortlessly.

Usage is the first catch. GrokBot seems to burn through its usage allowance quickly, and it's not very clear how much usage each price plan actually gives you.

The $30 entry plan probably gives you very limited usage in practice. A lot of people are reporting hitting their limits very quickly, so budget for a higher tier if you want to run bots all day.

James had a rather different week to Kieran. He really struggled to get GrokBot working at all, found it unreliable, and doesn't think it lives up to the online hype yet.

His view is that the model is genuinely interesting, giving every agent a full cloud computer and letting them talk to each other, but in day-to-day use it doesn't work reliably enough yet.

That should carry some weight. James is the more technical of the two of us and uses almost every tool out there, so if he couldn't get it running smoothly, plenty of others won't either.

The technology underneath GrokBot is not revolutionary. It's broadly what OpenClaw and Hermes already do, packaged better, and it's surprising what a difference a smoother UX makes.

Some people online say they've moved everything to GrokBot, including their coding. The underlying models are very similar, so that's less mad than it sounds, but James's week suggests we wouldn't move anything important over just yet.

If you're curious, the $30 a month entry tier makes it a cheap experiment, as long as you go in expecting to bump into the usage limits. Start with one bot doing one job, something like inbox triage, and build the office up from there.